This study reports development of a novel method for high-resolution in vivo imaging of the function of individual mouse retinal ganglion cells (RGCs) that overcomes many limitations of available methods for recording RGC physiology. The technique combines insertion of a genetically encoded calcium indicator into RGCs with imaging of calcium responses over many days with FACILE (functional adaptive optics cellular imaging in the living eye). FACILE extends the most common method for RGC physiology, in vitro physiology, by allowing repeated imaging of the function of each cell over many sessions and by avoiding damage to the retina during removal from the eye. This makes it possible to track changes in the response of individual cells during morphological development or degeneration. FACILE also overcomes limitations of existing in vivo imaging methods, providing fine spatial and temporal detail, structure-function comparison, and simultaneous analysis of multiple cells.
- retinal ganglion cells
- in vivo adaptive optics imaging
- calcium imaging
available methods for examining the physiology of retinal ganglion cells (RGCs) have marked limitations, which are resolved by the novel in vivo imaging method described here. Existing in vivo methods study cells in a relatively unperturbed state, but the physical challenge of recording within the eye often limits recording to a small number of cells (e.g., Martin et al. 2001), and repeated study of individual cells is not possible. In vitro recording is a highly efficient approach, imaging or recording electrophysiological responses from many cells simultaneously (e.g., Borghuis et al. 2011; Chichilnisky and Kalmar 2002), but it has two major limitations. It does not permit long-term study of the same retinal neurons and circuitry over time, for example, the study of how neurons degenerate or how they change during development. Also, in vitro methods do not avoid possible artifacts due to retina removal itself, for example, the effect of cutting the axons of ganglion cells (Lukas et al. 2009; Mandolesi et al. 2004).
Here we demonstrate an in vivo imaging method, FACILE (functional adaptive optics cellular imaging in the living eye), that can repeatedly image the light response of ganglion cells in the eyes of living mice. This method shares the advantages of other in vivo techniques in studying retinal cells in an intact state, and it can track the response of individual cells over days or weeks. Like other imaging methods, FACILE permits the study of many cells simultaneously, but unlike low-resolution imaging (Prilloff et al. 2010; Sabel et al. 1997), it provides both subcellular resolution and high sensitivity. Fluorescent adaptive optics (AO) imaging has previously been used to image the morphology of retinal ganglion cells expressing fluorescent protein in living rodent eyes and characterize their dendritic stratification (Geng et al. 2009, 2012). Here, to study ganglion cell physiology, we expressed the genetically encoded calcium indicator (GECI) G-CaMP3 (Tian et al. 2009) in retinal ganglion cells and tracked the calcium response of individual cells to visual stimuli. A challenge of in vivo fluorescent imaging is that it uses visible light that strongly light adapts the retina. For example, in this study the excitation wavelength was 488 nm, which is close to the peak sensitivity of mouse middle-wavelength sensitive (M) cone opsin and rhodopsin (Jacobs et al. 2004; Lyubarsky et al. 1999, 2004). To separate visual activation from response to the imaging light, we used ultraviolet (UV) light (365 nm) to stimulate the retina thus selectively activating the short-wavelength sensitive (S) cone opsin, which activates most retinal ganglion cells (Wang et al. 2011).
Because in vivo imaging is intrinsically photon limited, we used a viral-vector-mediated gene delivery that leads to a high level of G-CaMP3 expression in cells without damaging the cells. The rabies vector produced intense G-CaMP3 expression in individual ganglion cells without nonspecific background expression, which improved the signal-to-noise ratio of imaging. As described below, rabies mediated G-CaMP3 expression produced stable physiological responses in cells over the first 4–8 days of imaging, which then declined as the cells began to degenerate.
MATERIALS AND METHODS
Intracranial injection of rabies vector.
The rabies vector was produced in the laboratory of Dr. Edward M. Callaway (Salk Institute for Biological Study, San Diego, CA). In brief, G-deleted rabies viruses are generated in B7GG cells, concentrated by two rounds of centrifugation and titrated in HEK293t cells as described previously (Osakada et al. 2011). The titers of the viruses used in this study were 4.8–8.0 × 108 infectious units/ml. The injection procedure was as described previously (Cetin et al. 2006) and followed the guidelines of Biosafety Level 2. Typical coordinates of injections into superior colliculus were 3.8-mm posterior to Bregma, 1.0-mm lateral, at two depths of 1.4 and 1.7 mm (Keith and Franklin 2007), and the injection volume was 0.5 μl at each depth. During injections, the cornea was lubricated with GenTeal (Bausch & Lomb, Rochester, NY) to prevent cataract formation. AO imaging began 4 days after injection to allow recovery from surgery, as well as ganglion cell expression of G-CaMP3. Wild-type C57BL/6J mice of both sexes were used. All animal procedures were conducted according to the Association for Research in Vision and Ophthalmology Statement for the Use of Animals and the Guidelines of the Office of Laboratory Animal Care at the University of Rochester. The protocol was approved by the University Committee on Animal Resources of the University of Rochester.
Before AO imaging, the retinal distribution of transduced ganglion cells was identified on low-resolution fluorescent images of the mouse retina, obtained with a fundus camera (Micron III, Phoenix Research Laboratories, Pleasanton, CA) or confocal scanning laser ophthalmoscope (SLO; Spectralis HRA, Heidelberg Engineering, Heidelberg, Germany). FACILE was done with the mouse fluorescence AO SLO built at the University of Rochester. A detailed description of this system and characterization of its performance for imaging various retinal structures have been reported elsewhere (Geng et al. 2012). Briefly, during imaging, we captured two channels of images simultaneously: a reflectance image of the retinal structure (e.g., vessels in the inner retina), using infrared light with center wavelength 790 nm and a fluorescence image of the G-CaMP3 expressing ganglion cells (excitation: 488 nm and emission: 525 ± 17.5 nm; FF01–520/35–25; Semrock, Rochester, NY). Infrared light of 850 nm was also used for wavefront sensing to correct the optical aberrations of the mouse eye. All imaging lights, 488, 790, and 850 nm, were scanned over a 6 × 5° rectangular region on the retina, and AO images were taken over a central 5 × 5° square, where illumination was more uniform. The typical intensities of the imaging lights at pupil were ∼100 μW (488 nm), ∼150 μW (790 nm), and ∼25 μW (850 nm).
Eye motion was calculated from high signal-to-noise ratio reflectance images of retinal structure (e.g., vasculature), and the motion correction applied to both channels. Images of both channels were captured at 25 Hz. The gain of the fluorescence channel was set to fully utilize the dynamic range of the system, with only a small percentage of pixels being saturated. For navigation across the retina, the retinal vasculature visible in AO reflectance was compared with low-resolution, large field fundus images.
Presentation of UV light.
The primary visual stimulus was 365-nm UV light. The UV light was generated by an LED that produced a peak wavelength of 365 nm (M365L2-UV; Thorlabs, Newton, NJ) and was presented to the eye in Maxwellian view over a circular region of ∼24° diameter on the retina. The intensity of the UV light at pupil was 20 μW.
Visual impact of imaging and visual stimulation lights.
To compare the visual impact on photoreceptors of the various imaging and stimulation lights, we calculated isomerization rates for mouse M and S pigments and rhodopsin using methods described by Naarendorp (2010) (see also, Wang et al. 2011) (Fig. 1, A and B). The spectral sensitivities followed the template developed by Govardovskii et al. (2000) using λmaxs of 508, 360, and 498 nm (Jacobs et al. 2004; Lyubarsky et al. 1999, 2004). Pigment self-screening was calculated based on inner segment lengths of 13.4 um for cones and 24 um for rods measured by Carter-Dawson and LaVail (1979) and an axial specific density of 0.015 um−1 for cones and 0.019 um−1 for rods (Naarendorp et al. 2010). The end-on collecting area of mouse rods and cones were 0.85 and 1 μm2 (Naarendorp et al. 2010). Preretinal absorption of the mouse eye was estimated from Henriksson et al. (2010). To simplify the calculation, all light sources were treated as monochromatic light, with all the power at the spectral peak. To estimate the bleaching effect of imaging lights, we assumed a total 7.4 log10 pigments in a M cone (Fu 2011), estimated from cone outer segment volume of 14 μm3 (Carter-Dawson and LaVail 1979) and a pigment concentration of ∼3 mM (Harosi 1975). Psychophysics Toolbox extensions (Brainard 1997; Pelli 1997) were used for calculating isomerization rates. Although the UV stimulus light produced visible autofluorescence at the contact lens surface, the confocal AO system eliminated this fluorescence signal.
The primary stimulus used to activate ganglion cells was an 8-s steady presentation of 365-nm (UV) light, which alternated with 8 s of no UV light. For a few measures other stimuli were used; 1-s steady presentation of UV light, alternating with 15 s of no UV light or onset of the 488-nm imaging field.
To minimize retinal exposure to blue light, the 488-nm imaging light was extinguished between each imaging block. At each retinal location, an imaging block was of an ∼3-min duration, with at least 1 min between recording blocks. Baseline fluorescence generally decreased over the first 30 s of each block; thus data analysis excluded the initial 30 s of each block.
Measurement of fluorescence intensity of ganglion cell somas.
Fluorescent measurements were made from motion-corrected (registered) fluorescent videos, using Fiji (Schindelin et al. 2012), a distribution package of ImageJ (W. S. Rasband, ImageJ, U. S. National Institutes of Health, Bethesda, MD; http://imagej.nih.gov/ij/, 1997–2011). For each registered fluorescent video, round or ellipsoidal regions of interests (ROIs) were placed on each ganglion cell soma identified in the summed image of that video (see Fig. 2). As a measure of background, a rectangular ROI larger than a typical ganglion cell soma was placed at a location where there were no visible ganglion cell somas. Fluorescence intensity was measured within ROIs for each frame of the video, generating a temporal sequence of fluorescence intensity for each identified ganglion cell soma.
To improve the signal-noise-ratio of the measurement (as illustrated in Fig. 2B, left), we binned 10 successive samples, reducing the temporal sampling rate from 25 to 2.5 Hz, a rate that is fast relative to the calcium response of ganglion cells. A time stamp was recorded for every video frame during image capture to correctly align each frame with the timing of stimulus presentation.
Measurement of peak response from normalized fluorescence.
We calculated fluorescence response (ratio F/F0) by normalizing measured fluorescence (F) of each stimulus cycles to the mean baseline value (F0) obtained from the time preceding the onset of the UV flash (2–0.4 s before onset). Time of the peak response was identified from smoothed data with a three-sample moving boxcar average, and the peak response amplitude was determined as the mean of three consutive time points statistically different from baseline (F/F0 = 1), using a one-tailed t-test. All data analysis used MATLAB (MathWorks, Natick, MA).
Evaluation of light safety with repeated imaging.
Total light exposure included the imaging lights, primarily 488 and 790 nm, and the visual stimulation light, 365 nm. Exposure to the imaging lights did not exceed 1.7 times the American National Standards Institute (American National Standard for the Safe Use of Lasers ANSI Z136.1–2007) maximum permissible exposure (MPE) for human retina (Delori et al. 2007) scaled by the ratio of the numerical aperture (NA) of human to mouse eyes (NAHuman2/NAMouse2 = 0.2) (Geng et al. 2012). This exposure level was considered unlikely to be damaging, given that the MPE is set 10-fold below minimum damage thresholds to ensure safety. Maximum radiant exposure to UV light for each eye, each day was 0.18–1.02 J/cm2. This value cannot be compared with human safety limits since there is no MPE standard for UV light, but this light level is of concern, given that it approaches the value that has been shown to produce retinal damage in rat (Busch et al. 1999; Gorgels and van Norren 1995; van Norren and Gorgels 2011). Despite this concern, no photoreceptor alteration or loss was found in AO imaging at any retinal location on any of the imaging sessions over 7 days nor was there any visible change in cornea or lens. Our previous studies have demonstrated that AO imaging is a highly sensitive indicator of light damage, showing retinal changes in monkey retina at light levels well below ANSI MPE standards for light safety (Morgan et al. 2008).
Combining calcium imaging with in vivo fluorescent AO imaging.
This study combined a previously described method for high-resolution in vivo AO imaging of reflectance and fluorescence signals from mouse retina (Geng et al. 2012), with insertion of the genetically encoded calcium indicator G-CaMP3 into RGCs (Tian et al. 2009). G-CaMP3 was expressed in mouse RGCs by injecting glycoprotein-deleted rabies vector (Osakada et al. 2011), carrying the G-CaMP3 gene, intracranially into mouse superior colliculus, producing RGC transduction in a few days. The apparatus and method for AO imaging were as previously described (Geng et al. 2012) with the following exceptions. The imaging laser in the present study was 488 nm rather than 514 nm, chosen to match the excitation of G-CaMP3, and lower laser intensities were used in this study to maintain safe light levels during the extended imaging needed to characterize physiological responses (see materials and methods).
Visual stimuli were 488 nm (blue) or 365 nm (UV).
To excite retinal neurons we initially used the onset of the 488-nm (blue) imaging field, a wavelength near the peak sensitivities of both mouse M cones and rods (Figs. 1A and 2A). However, to minimize interactions between the stimulation and imaging light, most measurements in this study were made of responses to 365 nm (UV) light (Figs. 1A and 2B), obtained in the presence of constant 488-nm illumination. As shown in Fig. 1A, 365-nm light is near the peak sensitivity of mouse S cones (Jacobs and Rowe 2004; Lyubarsky et al. 1999; Nikonov et al. 2006) and activates both S cones and the many cones that coexpress M and S opsins (Applebury et al. 2000; Haverkamp et al. 2005; Jacobs and Williams 2007; Rohlich et al. 1994; Wang et al. 2011). To reduce the visual impact of the 488-nm imaging light on S cones, the intensity of the UV stimulus was set ∼3 log10 units above that of the 488-nm fluorescence imaging light (Fig. 1B).
Calcium responses to brief stimuli.
We examined the time course of calcium responses to brief stimuli, initially to onset of the 488-nm imaging field. Figure 2A illustrates the G-CaMP3 response of a ganglion cell in the living mouse eye to onset of the 488-nm imaging light. The response (blue curve) showed a quick rise to peak, followed by a slower decay to baseline within 10 s. This response demonstrates that the 488-nm imaging light activates retinal neurons, suggesting that calcium responses might best be understood if the 488-nm light was held constant. Figure 2A, right, shows the method used to determine calcium responses on a summed fluorescence image of ganglion cells expressing G-CaMP3. To determine the time course of G-CaMP3 fluorescence, we measured mean fluorescence intensity within oval ROIs at the selected ganglion cell soma (Fig. 2A, white oval at right).
To separate the effects of visual stimulation from activation by the imaging field, we kept the 488 imaging field constant and measured the calcium response to a brief pulse of UV light. Figure 2B shows the time course of the calcium response to a 1-s pulse of UV light, measured from the ganglion cell soma within the white oval ROI at right. The fluorescence intensity (Fig. 2B, red curve) rises slowly, peaking at ∼2 s and returning to baseline ∼4 s after stimulus onset.
Calcium responses to UV stimuli of longer duration.
Following the above measurement of responses to brief stimuli, the remaining data in this study were collected using an 8-s presentation of the UV stimulus, to reveal both transient and sustained calcium responses. Figure 3 shows examples of fluorescent responses to UV stimuli measured on day 6 following rabies injections. The temporal profile of responses varied across cells, increasing in some and decreasing in others. For cells having increased responses, the responses of some cells continued to increase or plateaued as long as the UV stimulus was present (e.g., Fig. 3, A, top, and B, top row), while responses of other cells peaked early during the 8-s UV stimulus and then declined (e.g., Fig. 3B, 1st, 3rd, and 4th panels from left in the 2nd row). For all cells with a decreased response, responses generally continued to decrease as long as the UV stimulus was present. However, after UV stimulus offset, the response of some cells recovered to baseline level rapidly (e.g., Fig. 3A, bottom, and Fig. 3B, middle in the 3rd row and the 1st panel from left in the 4th row), while the response of other cells recovered more slowly (e.g., Fig. 3B, 1st and 3rd panels from left in the 3rd row). To estimate the main peak response amplitude across ganglion cell population, we quantified peak responses of 60 imaged cells measured on day 6 (see appendix). Eleven out of the 60 cells showed statistically significant increase in fluorescence; 41 out of the 60 cells showed statistically significant decrease in fluorescence. The mean amplitude of increased fluorescence was 0.4 ± 0.4 (SD), which was approximately threefold greater than that of decreased fluorescence: 0.15 ± 0.04 (SD).
Calcium responses of ganglion cells across time after rabies injection.
To evaluate the value of FACILE methodology for chronic study of neuronal responses, we examined the calcium response of individual ganglion cells across multiple days during which the cells degenerated. Figure 4A shows repeated imaging on days 4, 6, 8, and 10 of two clusters of G-CaMP3 expressing ganglion cells in the retinas of two different mice. The total number of transduced ganglion cells was highest around day 6 (Fig. 4A, bottom). To assess the consistency of cell responses over multiple days, normalized responses of selected cells were shown in Figure 4B (see Fig. 3 also, for cells identified with same numbers). A transduced ganglion cell was typically visible for at least three of the time points, and different cells initially became visible on different days. For example, cells 1–4 were visible on days 4–8, whereas cell 5 was visible on days 6–10 (Fig. 4A). For cells showing increased fluorescence to visual stimulation, responses were typically stable for two time points and then greatly decreased at the third time point (e.g., cells 1–3 in Fig. 4B), whereas those showing decreased fluorescence often remained relatively stable until they could no longer be imaged (cells 7–12 in Fig. 4B).
This study demonstrated a novel method for the repeated in vivo imaging of calcium responses of individual retinal ganglion cells in the mouse eye. It utilized an AO imaging system (Geng et al. 2012; Gray et al. 2006, 2008) to obtain high-resolution images of fluorescing cells in retina despite the challenges of low fluorescence levels and eye movements. We employed an UV light source that activates the S-cone opsin to measure ganglion cell responses in the presence of the blue imaging laser, which also excites M-cone opsin and rhodopsin. Correction of optical aberrations by the AO system permitted imaging of the subcellular structure (soma, proximal dendrites) of ganglion cells (Fig. 2, A and B, right). Finally, the level of both imaging and stimulation light were kept low to avoid light damage that would compromise the repeated imaging of cells. This is the first report of high-resolution in vivo imaging of the light responses of individual retinal ganglion cells and shows imaging of calcium responses of ganglion cells at temporal and spatial resolutions comparable to those used in similar in vitro studies (Borghuis et al. 2011; Briggman and Euler 2011; Weitz et al. 2013; Zariwala et al. 2012). As described above, important advantages of the FACILE method are simultaneous, high-resolution in vivo study of numerous cells, repeated imaging of single retinal cells over a long time, and the ability to relate the physiology of recorded cells to their morphology. The high axial resolution of FACILE can distinguish RGC cell types on the basis of their branching level within the inner plexiform layer (Geng et al. 2012). A major limitation of FACILE imaging of calcium is poor temporal resolution compared with electrophysiology. However, in the future, we could avoid the inherently slow calcium response by using voltage-sensitive indicators (Mutoh et al. 2012). Other limitations of the present method can be reduced by incorporating methodologies described below in future directions.
Comparison with prior in vitro calcium imaging studies of RGCs.
The calcium response of ganglion cells to light stimulation has previously been recorded with in vitro preparations (Borghuis et al. 2011; Briggman and Euler 2011; Zariwala et al. 2012). These studies used either a calcium indicator dye, Oregon Green BAPTA, or a genetically encoded calcium indicator, G-CaMP3, imaged with a two-photon laser. Briggman and Euler (2011) studied the calcium response of mouse ganglion cells to a 1.5-s presentation of a 420- and 580-nm spot that ranged in diameter from 50 to 800 μm. Responses to a 800-μm spot (their Fig. 2A), the condition most similar to that used in the present study, showed increased or decreased fluorescent response similar to that seen in Fig. 3 in the present study. Borghuis et al. (2011) studied the calcium responses of mouse ganglion cell to a 2-s presentation of a full-field 458-nm stimulus (their Fig. 6B) and also showed increased or decreased responses to light in some cells similar to that seen in Fig. 3 of this study.
Both Borghuis et al. (2011) and Briggman and Euler (2011) showed cells with clear OFF responses, i.e., fluorescence increase above baseline level to light off. While many cells in our study showed decreased responses during the 8-s UV flash presentation (Fig. 3), none showed increased fluorescence above baseline after the 8-s UV flash was extinguished. The ∼24° UV spot used in our study was much larger than the optimal 10° spot diameter for mouse ganglion cells (Schmucker and Schaeffel 2004; Stone and Pinto 1993), which can greatly reduce OFF responses of OFF-center ganglion cells (Sagdullaev and McCall 2005). Thus, although our responses appear consistent with those reported by Briggman and Euler (2011) for spot diameter of 800 μm, smaller spot sizes will be needed to demonstrate unequivocal OFF responses.
Future directions for FACILE.
The FACILE method described here can be extended to additional applications by slight changes in calcium indicators, visual stimuli, imaging methods, and even different species. Long-term retinal changes could be examined over weeks to months by the use of less damaging methods for inserting GECIs into retinal cells than the rabies virus used here. Precisely focused patterned stimuli, generated with the AO system, can be used to map receptive fields with high precision. The method is currently being extended to nonhuman primates, which offer an animal model of retinal function and dysfunction that is closely related to human retina. Different imaging modalities will greatly extend the capabilities of the FACILE method. For example, the use of fluorescence resonance energy transfer (FRET)-based sensors can provide both ratiometric calcium imaging and structural imaging and extend the method towards studying fine scale molecular interactions. Furthermore, multiphoton methods (Hunter et al. 2011) will permit noninvasive FACILE imaging of many compounds involved in the retinoid cycle, thus potentially extending the approach to the human retina.
This work was supported by National Eye Institute Research Grants EY-019375, EY-021166, EY-004367, BRP-EY-014375, and NDC-5PN2EY-018241; National Eye Institute Training Grant EY-07125; and National Eye Institute Core Grant-EY-001319.
D. R. Williams has license agreements for AO technology with Canon and Physical Sciences and consults for GlaxoSmithKelin and Pfizer. He also has a Small Business Innovation Research grant on development of novel AO technology with Polgenix. All other authors declare no competing financial interests.
Author contributions: L.Y., E.M.C., D.R.W., and W.H.M. conception and design of research; L.Y., Y.G., F.O., R.S., A.H.C., and W.H.M. performed experiments; L.Y., D.R.W., and W.H.M. analyzed data; L.Y., D.R.W., and W.H.M. interpreted results of experiments; L.Y. and W.H.M. prepared figures; L.Y., D.R.W., and W.H.M. drafted manuscript; L.Y., Y.G., F.O., R.S., A.H.C., E.M.C., D.R.W., and W.H.M. edited and revised manuscript; L.Y., Y.G., F.O., R.S., A.H.C., E.M.C., D.R.W., and W.H.M. approved final version of manuscript.
We thank Jennifer Strazzeri and Stacy Ruvio for assistance with surgery; Jennifer Norris for assistance with data processing; William Fischer, Rachel Hollar, and Lee Anne Schery for assistance with cSLO or AO imaging; Kamran Ahmad for assistance with imaging software programming and hardware interface; Jennifer Hunter for assistance with light safety calculations; and Tracy Bubel for assistance with histology. Loren Looger supplied the G-CaMP3 plasmid and provided advice on its use.
Quantification of calcium responses across the RGC population.
On day 6 after rabies injection, we imaged the calcium responses of total 60 transduced cells from 6 retinal locations in 5 mouse eyes. Figure A1 summarizes the measured peak response (F/F0) and time to peak across population.
- Copyright © 2013 the American Physiological Society